The Impact of File Sharing on Movies Koleman Strumpf University of Kansas [email protected] Felix Oberholzer-Gee Harvard University [email protected] 1. Introduction The advent of file-sharing technology makes it possible for consumers to copy music, books, video games and other protected works on an unprecedented scale at minimal cost. By some estimates, file sharing now accounts for up to sixty percent of all bandwidth usage (Sandvine, 2002-2007 and 2008ab). Music consumption was first affected by the new technology, mostly because the small size of music files allowed for fairly rapid transfers even in the early days of file-sharing when the technology was often slow and unreliable. By contrast, consumers gained the ability to share the much larger files associated with movies only with the emergence of BitTorrent-based file-sharing services. With the help of the BitTorrent technology, consumers can now download a feature film in less than two hours. BitTorrent is also significant because the technology forces users to share the parts of files that they already own while they download the remaining bits. This procedure limits the opportunity to free-ride that plagued older file-sharing systems. In view of these advances, it is no surprise that movies are now among the most heavily traded works on file-sharing sites. Academic research on the impact of file sharing on commerce and consumption mirrors the overall technological development. We know most about the effect of file sharing on music – while estimates vary, a good number of studies seem to indicate that file sharing is perhaps responsible for no more than a fifth of the decline in music sales in recent years (Oberholzer- Gee and Strumpf 2009) – but we have comparatively little knowledge about the influence of file sharing on movie consumption. In this paper, we take a first look at the influence of file sharing on the attendance and revenue of movie theatres. The idea is to see whether the extent of file sharing prior to a theatrical release is systematically related to the demand for theater tickets. 1 Our case study is the movie “X-Men Origins: Wolverine,” which was available on file-sharing services more than four weeks prior to its theatrical release. We exploit this unusual leak – unusual both because the film was available far earlier than is typical and in a quality that closely resembled the eventually released legitimate copy – to determine the influence of file sharing on the demand for theater tickets under these adverse circumstances. 2. Background – File Sharing File sharing refers to the computer protocols which facilitate the free transfer of digital files between users. On such networks copyrighted material is often transferred without the consent of or compensation paid to its owner. The first such file sharing networks, such as Napster and KaZaA, specialized in transfers of music files. Starting around 2004, BitTorrent became the dominant file sharing network in the US. BitTorrent facilitates the transfer of large files, which are broken into smaller chunks that are swapped using an efficient tit-for-tat rule. This new technology led file sharing users to focus on movies: a recent estimate is that between half and three-fourths of BitTorrent transfers involve video (Torrentfreak, 2010). File sharing has also become relatively mainstream, with an estimated 60m users in the US (BigChampagne. 2010).1 Downloading a file on the BitTorrent network requires two steps. First a user visits an index site which provides links to the files which are available for downloading. Index sites can be searched and accessed using a typical web browser. Some of the more well known indexers include The Pirate Bay, Isohunt, Mininova, Demonoid, and KickAssTorrents, and the links on these sites often overlap. In the second step, the index links point to a cloud of users who are currently uploading and downloading the particular file. A computer called a tracker manages each cloud (a single tracker typically manages many such clouds, and many clouds have multiple trackers). A user’s experience on BitTorrent varies based on some technological features. Upload or download speeds can vary based on local internet congestion, which changes over time based on technical issues (severed fiber lines or ISPs adding additional capacity) as well as demand (local internet activity, including those unrelated to file sharing). The BitTorrent infrastructure also plays a role. Index sites and trackers can crash or become faster, often times for reasons unrelated to demand (for example, index sites may remove all unauthorized content or suddenly open following an extended dormancy). Legal actions also matter, as both index sites 1For a more detailed timeline of file sharing, see Oberholzer-Gee and Strumpf (2009). 2 and trackers can disappear due police raids or court orders (the latter can go into effect long after an initial verdict or when site owners decide to stop making appeals). 3. The Movie Industry The US movie industry derives revenues from a wide range of sources. While there is readily available data for theatrical exhibition and home video (which includes sale and rental of DVDs, BluRay Discs, video on demand and electronic sell-through), there is little information for ancillary streams such as television licenses.2 Figure 1 presents aggregate real revenues for the US market over the last decade. While theatrical revenues are steady, home video revenues rose in the beginning of the decade and declined in the second half. The latter is important since home video revenues are roughly twice as large as theatrical revenues. Figure 2 shows theatrical revenues for foreign markets.3 Non-US revenues have doubled in the last decade and now rival the US home video market in size.4 A couple of general points emerge from these data. The movie industry enjoyed strong growth in the early part of this decade, but revenues have been flat (theatrical) or even declining (home video) over the past few years. Since this coincides with the growth of video file sharing, some have argued this is evidence of a crowd-out effect. While a possibility, there are reasons to be skeptical these time series patterns are due to a causal relationship. First there are several alternative factors which could explain the recent revenue declines. For example part of the previous growth was fueled by home consumers starting up a library of the new format (DVD), and as the format matured this effect dissipated (real DVD revenues increased steadily through the early part of the last decade, peaked in 2006 and then fell 14% over the next three years).5 Also during this period there large changes in the rental market with the rise of Netflix and Redbox. At the same time, no new format has caught on in a 2Based on confidential MPAA documents, Epstein (2005, 2006) states that television licensing revenues from cable, traditional networks, syndication and foreign broadcasts nearly match those from home video in 2004, and are far more profitable due to lower costs. Vogel (2007) presents comparable numbers for 2005 and of specific movies. 3There are no public data for home video revenues in foreign markets, but Epstein (2005, 2006) reports they were about two-thirds the size of the US home videos. 4Vogel (2007) has historical time series of revenue and costs within the industry. 5While home video sales exceed theatrical revenues, the most popular titles make more money from ticket sales than from DVD sales. For example, in 2009 the top 100 theater titles sold $9.4b in tickets while the top 100 selling DVD titles had revenues of $4.0b (there is a high degree of overlap in the titles on these two charts). It is the catalog titles, which have little or no theatrical sales, which explain why the home market is larger. 3 significant way to replace the decline of the DVD. Second, while the domestic movie industry has struggled, the experience of the international market is more mixed or even positive. Since file sharing is a global phenomenon, if it crowds out sales in the US it should also be doing so in other countries. All this suggests that a more detailed analysis is needed to quantify the displacement of file sharing on video markets. 4. The Wolverine Experiment Movies typically become available on file sharing networks within a week or so following their release in theaters (Figure 3 shows the typical release pattern for the top 150 grossing movies each year between 2005 and 2009). These initial copies are typically low quality and come from video cams made in the theater. A rare exception to this case occurred in April 2009 when a copy of the movie “X-Men Origins: Wolverine” was made available a full month before its release in theaters.6 In addition to being available so early, the copy was a workprint which is virtually identical to the copy shown in theaters (minus some final special effects). The movie was downloaded about 1m times on BitTorrent in its first week of availability, 4.5m times by the time it was released in theaters, and 7.2m times over all 2009. A majority of these downloads were made outside the US. The Wolverine case provides a unique opportunity for studying the impact of file sharing on movies. First, it provides a worst-case scenario in which a high quality copy of the movie was readily available prior to its theatrical showing. The extensive media coverage of the movie’s leak probably meant virtually of the potential theatergoers were aware they could download it in advance. If individuals use file sharing to substitute for paying to see a film, it should be apparent in this case.